提交 · 8965819fbbd23f5ff159903eb673898c45408aa4 · PaddlePaddle / PaddleDetection

21 3月, 2019 1 次提交
- Z
  
  rewrite the cuda kernels of channel_wise_quant_op and channe_wise_dequant_op. test=develop · 8965819f
  由 Zhen Wang 提交于 3月 21, 2019
  
  8965819f
19 3月, 2019 1 次提交
- Z
  
  add channel wise quantization in ir pass. · ec88b6cc
  由 Zhen Wang 提交于 3月 19, 2019
  
  ec88b6cc
15 3月, 2019 3 次提交

add moving average absmax op and fix bug (#15155) · 81b4fad8
由视言提交于 3月 15, 2019
```
* Add moving average absmax op in quantilize-aware training.
```
81b4fad8

Support sync batch norm. (#16121) · 8ad672a2

由 qingqing01 提交于 3月 15, 2019

* Support Sync Batch Norm.
* Note, do not enable it in one device.

Usage:

build_strategy = fluid.BuildStrategy()
build_strategy.sync_batch_norm = True
binary = fluid.compiler.CompiledProgram(tp).with_data_parallel(
        loss_name=loss_mean.name,
        build_strategy=build_strategy)

8ad672a2

Y
Impl fp16 compute kernel for slice_op (#16206) · 4ae23cc3
由 Yibing Liu 提交于 3月 15, 2019
```
* Impl fp16 compute kernel for slice_op

test=develop

* Use data() to replace mutable_data()
```
4ae23cc3

14 3月, 2019 6 次提交
- S
  fix numeric error · 3e036956
  由 sneaxiy 提交于 3月 14, 2019
```
test=develop
```
  3e036956
- S
  revert revert 16144 · 5a92e4c0
  由 sneaxiy 提交于 3月 14, 2019
```
test=develop
```
  5a92e4c0
- S
  fix numeric error · 55ba7f61
  由 sneaxiy 提交于 3月 14, 2019
```
test=develop
```
  55ba7f61
- S
  include unordered_map to cross_entropy_op.cc · e993effb
  由 sneaxiy 提交于 3月 14, 2019
```
test=develop
```
  e993effb
- Z
  Revert "PaddingRNN model memory optimize" · a91964c8
  由 Zeng Jinle 提交于 3月 14, 2019
```
test=develop
```
  a91964c8
- L
  1. disable reuse SELECTED_ROWS type variable (#16150) · 1c6caf84
  由 liuwei1031 提交于 3月 14, 2019
```
2. remove lod check in reshape op
test=develop
```
  1c6caf84
13 3月, 2019 4 次提交
- S
  fix travis-ci · 487624e1
  由 sneaxiy 提交于 3月 13, 2019
```
test=develop
```
  487624e1
- G
  resolve #15618 (#16114) · decdbed0
  由 guomingz 提交于 3月 13, 2019
```
* resolve #15618
Backgroud: the PR #15398 raised the box_coder op performance regression, we optimized the code via the more efficency leveraging opemmp.
```
  decdbed0
- S
  combine op files · 1e9fd407
  由 sneaxiy 提交于 3月 13, 2019
```
test=develop
```
  1e9fd407
- B
  remove const_cast and refactor ngraph engine code (#15925) · e3c37bd5
  由 baojun 提交于 3月 13, 2019
```
* remove concast_cast and refactor code test=develop

* reduce flag use test=develop
```
  e3c37bd5
12 3月, 2019 12 次提交
- L
  refine with comments · fe78a92e
  由 luotao1 提交于 3月 12, 2019
```
test=develop
```
  fe78a92e
- D
  
  no use _gt_score. test=develop · d31693af
  由 dengkaipeng 提交于 3月 12, 2019
  
  d31693af
- L
  add runtime shape for fuse_emb_seq_pool_grad · 5d20954a
  由 luotao1 提交于 3月 12, 2019
```
test=develop
```
  5d20954a
- D
  
  add doc for param default. test=develop · aad62eec
  由 dengkaipeng 提交于 3月 12, 2019
  
  aad62eec
- T
  refine doc · a8e375d4
  由 tink2123 提交于 3月 12, 2019
```
test=develop
```
  a8e375d4
- D
  
  fix spell mistake in doc. test=develop · 585766ac
  由 dengkaipeng 提交于 3月 11, 2019
  
  585766ac
- D
  
  fix format. test=develop · b307533b
  由 dengkaipeng 提交于 3月 08, 2019
  
  b307533b
- D
  
  fix doc.test=develop · afdf3c3f
  由 dengkaipeng 提交于 3月 05, 2019
  
  afdf3c3f
- D
  
  fix API.spec not add defaults. test=develop · af4ef80e
  由 dengkaipeng 提交于 3月 05, 2019
  
  af4ef80e
- D
  
  fix unittest for yolov3_loss. test=develop · 0d1a9996
  由 dengkaipeng 提交于 3月 05, 2019
  
  0d1a9996
- D
  
  add mixup score and label_smooth for yolov3_loss. test=develop · f0804433
  由 dengkaipeng 提交于 3月 05, 2019
  
  f0804433
- S
  refine code · b26e9bd2
  由 sneaxiy 提交于 3月 12, 2019
```
test=develop
```
  b26e9bd2
11 3月, 2019 10 次提交
- L
  add all_kernels_must_compute_runtime_shape example for speedup infershape · 31ccaf09
  由 luotao1 提交于 3月 11, 2019
```
test=develop
```
  31ccaf09
- T
  enhance jitkernel unit test · 14d87112
  由 tensor-tang 提交于 3月 11, 2019
```
test=develop
```
  14d87112
- C
  Revert "Revert "Add Event for TensorCopy"" (#16035) · ad80bde8
  由 chengduo 提交于 3月 11, 2019
```
* Revert "Revert "Add Event for TensorCopy" (#16022)"

This reverts commit e2da3a5b.

* use default stream
test=develop
```
  ad80bde8
- T
  refine jitcodekey and enhance unit tests · cfc83c14
  由 tensor-tang 提交于 3月 11, 2019
```
test=develop
```
  cfc83c14
- S
  add unittest · cfd012e2
  由 sneaxiy 提交于 3月 11, 2019
```
test=develop
```
  cfd012e2
- S
  refine cross_entropy mem · d7407c90
  由 sneaxiy 提交于 3月 08, 2019
```
test=develop
```
  d7407c90
- S
  disable gc in recurrent_op currently · 732fa00e
  由 sneaxiy 提交于 3月 08, 2019
```
test=develop
```
  732fa00e
- T
  fix format · e4e0d034
  由 tink2123 提交于 3月 11, 2019
```
test=develop
```
  e4e0d034
- T
  Update activation_op.cc · 5579fae1
  由 Tink_Y 提交于 3月 11, 2019
```
test=develop
```
  5579fae1
- T
  enhance the jitkernel helper and add unit tests · 45bdd84d
  由 tensor-tang 提交于 3月 10, 2019
```
test=develop
```
  45bdd84d
10 3月, 2019 1 次提交
- T
  Add the inverse trigonometric function · 837ad7f8
  由 tink2123 提交于 3月 10, 2019
```
test=develop
```
  837ad7f8
08 3月, 2019 2 次提交
- T
  simplify the jitkernel templates and tests · 14a764c9
  由 tensor-tang 提交于 3月 08, 2019
```
test=develop
```
  14a764c9
- Y
  Make parent_idx a dispensable output for beam_search op to support models... · 66ead07e
  由 Yiqun Liu 提交于 3月 08, 2019
```
Make parent_idx a dispensable output for beam_search op to support models saved by older paddle version. (#16106)

test=develop
```
  66ead07e

PaddlePaddle / PaddleDetection 大约 1 年 前同步成功

PaddlePaddle / PaddleDetection
大约 1 年前同步成功